A Complete Overview of Word Embeddings
Вставка
- Опубліковано 27 тра 2024
- NLP has seen some big leaps over the last couple of years thanks to word embeddings but what are they? How are they made and how can you use them too?
Let's answer those questions in this video!
Get your Free Token for AssemblyAI Speech-To-Text API 👇www.assemblyai.com/?...
▬▬▬▬▬▬▬▬▬▬▬▬ CONNECT ▬▬▬▬▬▬▬▬▬▬▬▬
🖥️ Website: www.assemblyai.com
🐦 Twitter: / assemblyai
🦾 Discord: / discord
▶️ Subscribe: ua-cam.com/users/AssemblyAI?...
🔥 We're hiring! Check our open roles: www.assemblyai.com/careers
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
#MachineLearning #DeepLearning
Thank you. It is very clear and informative, though i really think you (AssemblyAI) should lose the music on the background; it is distracting and it gives the whole thing an infomercial feeling.
Somehow the music had a motivational influence for me. I caught myself vibing to it a few times
Would love a video on ELMo further. Thanks for all this!
There are maybe 30 videos on this topic and this is the only one that does not suddenly make a massive jump across whole concepts that the presenter knows but the watcher does not.
amazing video. Perfectly clear speech, good explanations, logical visualisations and the background music makes it a lot easier to focus. Thank you!!
Awesome overview.. Loved it.. Waiting for videos explaining GloVe and Elmo..
Great to hear you liked it!
great explanation. Please explain ELMO and GloVe. it was really great
Thank you for the suggestions!
@@AssemblyAII'd love to see those videos too
Would be great to see a video on Elmo!
Thank you for the suggestion, noted!
Excellent ! Thank you so much for making an absolutly clear explanation.
great explanation. please explain elmo and other approaches. also please make a video about efficient ways of clustering the embeddings👍
Thank you Sajjad for the suggestion!
Great explanation! I went through the topics hours of hours. But this channel saved my time. And on target.
Great to hear!
Excellent explanation. I did some study on this topic before coming here and the reason was because so many terms and concepts were quite overwhelming. I generally understood those but still missed the fine tuned clarity. After watching this video, most of what I read before started making a lot of sense. I highly recommend this video. Thank you so much.
This is great to hear! You are very welcome!
Very nice explanation of embedding concept, Would love to see pre-trained word embeddings for sentiment analysis.
Wow such a good presenter. I really like the examples super clear. This stuff is amazing
The absolute best video I've seen on this topic!!
Amazing Content.. Exactly what a learner wants .. to Have all the concepts in a single Video with easy to understand way in minimum time..
Thanks for taking the time to break this down and share!
You are very welcome! - Mısra
Very interested in an in depth explanation of ElMo
simple and clear explanation. please explain Elmo, thanks
Great explanation! Thanks for sharing
Very good explanation, thank you!
Thanks! Great information in a very objective way!
Yes - to all videos you suggest making! Great guide thank you.. was struggling to see value in lemmatization and concerned a bout a loss of coherence. Seeing several worked examples are great. Interested how the final results were all different but all had similarly high percentage match. How do you tackle this?
Great explanation in less amount of time. Really liked the video.
That's great to hear!
çok teşekkürler, bu kadar iyi anlatan başka video yok
Very clear, thank you
Great and very illustrative video
Interested in “Creating your own embedding before doing binary or multi label classification prediction”! Thanks for the clarity.
Very well explained!! Thank you so much
You're welcome!
Well explained ! Thanks a lot
Excellent presentation. I will be teaching this topic to students shortly and will recommend this material.
Great to hear, thank you!
Brilliant video, as always, thanks so much. Would love to see your suggested follow on using pre-trained word embeddings for sentiment analysis if you ever have time 🙂
This was awesome. Would love to see Elmo video and sentiment analysis video you mentioned possibly making!
amazing video!!!❣❣❣ Thanks for sharing
Very Good video. I second the other comments. PLEASE drop the music completely. It would increase the quality of the experience by at least 70%. I had hard time finishing the video because of the music
Thanks, will do!
Thanks dear. Nicely paced intro. Good for recap.
Glad you liked it
Thanks for the video! I've enjoyed watching and liked the format and pace. I'd add the retrowave background to my playlist if I knew the name. I guess that people would note it less if the volume was lower.
Great video. Thanks for sharing it. It would be great if you do a task like train sentiment analysis model with word embedding and share with us.
Great videos there, thank you for your content and keep up the good work!
Thank u very clear. Need to know how to use word embedding for text classification
Thank youuuu it's my first video but I guess I should make your video my periorties I'm NLP thanks alot❤
It's a really good explanation, thank you very much :)
You are welcome!
Awesome video!!
top video for embedding introduction
Great job 👍
Thanks that helped a lot.
Glad it helped
Thank you
thnak you soo much, amazing explaination and you beautiful
Clear explanation! 👍
Glad you think so!
nice video on word embedding keep it upp.............
Thank you!
Thanks for the video I do have a question when you said that for instance in the CBOW there is only one layer it means that the ouput of this layer should be a vector of size dimension of the embedding but in order to train the model we need to compaire this output with the word in the midlle which is actually a one hot encoded vector of size dimension of the vocabulary so it migth have another layer and a softmax.
I'm your fan already, please make an ELMo video....!!!
Thank you for the suggestion!
Great explanation
Thank you!
From the embeddings of your name, I removed those of "work", added "great" and "relationship" and I came up with the embeddings of my own name? How come? Mere coincidence? 🤔🤔
Great video, btw!
Thanks for the explanation please try to make a video about how ELMOS works
Great! Thanks
You're welcome!
This is amazing. Can you share the python notebook you show at 12m33s?
Video on Training a sentiment analysis model please
Great work !!
Can you make a video on Elmo and Transformer-based word embeddings ???
Great suggestion!
Excellent Explanation. I have one question please how could I fit my model with this embedding vectors cause for Example in one of my projects for extracting informations from fils. instead of using texts for training my models I thinked of using embedding but I don't know the best way to represent them to my model . I hope u understand my question and thank you.
Great video! As for your analogy, I would guess that changing cocktail to bar would indeed give you cocktail. The analogy of having dinner at a restaurant, is not matching to having bar at cocktail.
I love that all your examples are Lord of the Rings quotes because I run the Digital Tolkien Project which applies computational text analysis techniques to the works of Tolkien :-)
That's amazing! Nice to meet you! Huge Tolkien fan here. :)
@@AssemblyAI you should join the Digital Tolkien Project!
@@AssemblyAI Pls provide the notebook code ..
thnx
Great tutorial. She speaks like a native speaker. She looks like a Turkish girl, beautiful one :)
Would it be possible to use word embedding to ask if a text is about a certain topic (or rather to what degree a text is about a topic)?
Hi.. thank you for the video.. great introduction and also a practical example.. One request is to drop or reduce the intensity of the music. It was distracting.
Noted! Thank you for the feedback Praveen
Yes, great video but music is definitely too loud and distracting! It's really hard to concentrate on what you're saying.
I have just created my own word embedding algorithm (no neural networks). I am now training it. Let's see what kind of gibberish sentence it produces after I use it to produce new sentences (I will try to produce sentences without neural networks).
Here are some of the sentences produced by my word embedding (and just a word embedding without much on top of the word embedding).
"How nevertheless she had credited yourself."
"I released the chimneys were committed."-Well, this is two sentences "I released the chimneys" and "The chimneys were committed." The word embedding is not full NLP so this sort of word embedding cannot remember "released" when we get to "were".
"Luke fills the intruder boxed the interval."-Same issue here.
"I shall be linked unarmed."
"He was afraid I know."
be great to see a video on Elmo.
How large should data for a custom embedding be and is it possible to utilize a GPU for the creation of a word embedding vector space?
Can the embeddings from Transformer be used elsewhere, like with Word2Vec?
Great!
Thank you!
Is there a sentiment training model video that builds from this? Trying to build a recommendation system based on candidate sentences and a job description
We don't have that video yet but thank you for the suggestion!
8:15 i am having problem with the sentence "no of neurons in hidden layer = size of embedding".
i am confused what is size of embedding?
Great visual, Great Voice , Good pace of presentation . Everything is awesome in this video.
thanks for sharing :D
Thank you for the nice words Soheil! Glad it was helpful!
is there a video about sentiment analysis yet?
Be interested in seeing a python example of Word2Vec.
super helpful, but is there a version of this without the music?
Sorry about that! We got a lot of feedback in this. Let me see if we can upload without the music. :D
if we have a sentence "vishy eat bread". then we vectorize the word "eaat"(misspelled word), why does fasttext see that the word "eaat" is more similar to the word "eat"?. How is the architecture?, is it possible for fasttext without using skipgram to be able to classify words?. Thanks
New crush added to life
Great content thanks. Due to a hearing problem I would appreciate it, if you could remove the backround music. Ok? Thanks
How do I know which embedding will be best choice for a specific use case? How do I know which distance measure will be best?
Depends on your use case, cuz lets say if your use case contains more in general words like tea, king, actor, etc. then you may try different embeddings and see for yourself which ones are working well for particular examples from your use case
OR
If your use case is quite specific, something like say representing skills as a vector then you may need to train your own word2vec model on your data since pretrained embeddings may not cover what you need
Noice !
i would surely like to learn elmo guessing that chatgpt used the same correct me if i'm wrong 🙇🏻♂️
You should also add the name of the speak to videos. She says I in the video and we even do not know who is she :)
Waiting for the ELMo video.
Do transformers from scratch. I heard they can be written in 50 lines. I would like to understand how bert encodes words
Another good video marred by the inclusion of unnecessarily loud music.
Your pretty face holds my concentration, and thus I understand anything taught by you, especially transformer, more than any other youtube video..Thank you so much for such videos...indebted!
Thank you. Easy to understand. But I don't need the music at all. I fight myself listening to the music than your talk.
Can you make a video about ELmo?
Noted!
nice and crisp, just one suggestion "please remove background music", It is reductive to the viewers experience :)
Thank you! And noted!
Awesome content but these background music are slightly distracting specially when you play video on 1.5 speed
Cosine Similarity, not equal distance bro it just tells direction of that word
what about BOW?
Hi, Can you please tell you name. Going forward to learn more from you.
So when are we going to construct word embeddings from good old fashioned pictographs?
Why is there a background soundtrack during the lecture? Does it help with learning or focus? I find it kinda distracting and feel rushed.
Thanks for sharing! It would have been great to remove the background music.
Stoooooo[pppppppppp the awwwwwfffffuuuuulllll music!!!!! It’s beyond disracting😊
I was hoping this video would cover BERT as it can be used to generate embeddings. Bidirectional Encoder Representations from Transformers (BERT) is a family of language models introduced in 2018 by researchers at Google. However I do see there is another video about BERT: ua-cam.com/video/6ahxPTLZxU8/v-deo.html
excellent tutor but music distracted me so much 😄
Sorry about that! Wish we could take it back 🤷♀️
@@AssemblyAI I personally like both
From this lecture, ua-cam.com/video/OATCgQtNX2o/v-deo.html at 0:23, my understanding embedding is produced from the transform encoder. But from your lecture, the embedding is before transformer at 6:45. I also notice that in your lecture, the embedding is for a single word, but in this HF lecture, it seems for a whole sentence. I am still confused. Sorry that my question may be naïve.